On Not Having an Advance Abyssal Plan

by Eliezer Yudkowsky 11y23rd Feb 200943 comments

21


"Even though he could foresee the problem then, we can see it equally well now.  Therefore, if he could foresee the solution then, we should be able to see it now.  After all, Seldon was not a magician.  There are no trick methods of escaping a dilemma that he can see and we can't."
        -- Salvor Hardin

Years ago at the Singularity Institute, the Board was entertaining a proposal to expand somewhat.  I wasn't sure our funding was able to support the expansion, so I insisted that - if we started running out of money - we decide in advance who got fired and what got shut down, in what order.

Even over the electronic aether, you could hear the uncomfortable silence.

"Why can't we decide that at the time, if the worst happens?" they said, or something along those lines.

"For the same reason that when you're buying a stock you think will go up, you decide how far it has to decline before it means you were wrong," I said, or something along those lines; this being far back enough in time that I would still have used stock-trading in a rationality example.  "If we can make that decision during a crisis, we ought to be able to make it now.  And if I can't trust that we can make this decision in a crisis, I can't trust this to go forward."

People are really, really reluctant to plan in advance for the abyss.  But what good reason is there not to?  How can you be worse off from knowing in advance what you'll do in the worse cases?

I have been trying fairly hard to keep my mouth shut about the current economic crisis.  But still -

Why didn't various governments create and publish a plan for what they would do in the event of various forms of financial collapse, before it actually happened?

Never mind hindsight on the real-estate bubble - there are lots of things that could potentially trigger financial catastrophes.  I'm willing to bet the American government knows what it will do in terms of immediate rescue operations if an atomic bomb goes off in San Francisco.  But if the US government had any advance idea of under which circumstances it would nationalize Fannie Mae or guarantee Bear Stearns's counterparties, this plan was not very much in evidence as various government officials gave every appearance of trying to figure everything out on the fly.

A published, believable advance plan for the worst case - one that you could actually believe the government would carry out, instead of junking the plan to try to keep the top spinning a little longer - would have made the markets that much less uncertain.

If you don't publish a plan for catastrophe - or can't publish a believable plan - then the market just tries to guess what a realistic, believable plan would look like.  If that realistic, believable plan involves frantically attempting to bail out the large financial entities in order to keep the whole system from melting down further, you have moral hazard.  If they actually do it, that's lemon socialism (privatized upside, public downside).

If that's what happens in the abyssal case - then not publishing that fact, doesn't prevent anyone from foreseeing it.  If you publish that plan, maybe it will start a debate about whether to break up Bear Stearns into smaller entities, or change the plan to give counterparties a predictable 10% haircut, or claw back executive bonuses no matter what their contracts read (because you really aren't supposed to screw up so badly that the government has to get involved)...

But if you can't publish a realistic, believable advance abyssal plan that doesn't call for rescuing the huge entities - then who are you even kidding?

Governmental agencies failing to stare into the abyss in advance gives us a double problem: moral hazard as counterparties and investors try to guess what the government will realistically do; and fear and uncertainty in the market when the worst does happen.

It's questionable whether the government should be in the position of trying to forecast the abyss - to put a probability on financial meltdown in any given year due to any given cause.  But advance abyssal planning isn't about the probability, as it would be in investing.  It's about the possibility.  If you can realistically imagine global financial meltdowns of various types being possible, there's no excuse for not war-gaming them.  If your brain doesn't literally cease to exist upon facing systemic meltdowns at the time, you ought to be able to imagine plausible systemic meltdowns in advance.

Sure, you might have to make some modifications on-the-fly because you didn't get the exact causes and circumstances right.  But it shouldn't be obvious and predictable that the modifications will consist of "Oh dear it's more awful than we planned for and the systemic hazard is worse and now we really do have to bail out everyone even though we said we wouldn't."  Then the plan is not believable.

So long as the plan is not wrong in the stupidly obvious directions, it's hard to see how we'd be worse off if the governors of the Federal Reserve had taken a week once per year to play through scenarios more nightmarish than this one in their minds, deciding in advance what to do about it, realistically.

I suppose the main argument against publishing the plan would be that the uninformed public (i.e. Congress) would revolt against the emergency plans, demanding that unbelievable plans be substituted (let the banks burn! don't bail out GM!) and then changing their tune as soon as the worst actually happened.

But at least having the Federal Reserve privately visualizing all sorts of hideous possibilities in advance, war-gaming them with the Board of Governors, and planning for them realistically - so that when the worst starts happening, you don't have everyone running around being vague and visibly unprepared and refusing to talk about what happens if things get even worse - instead you just take out folder #37-B and figure out what needs tweaking - for the lack of that preparedness, there seems to me to be very little excuse.  The Federal Reserve should not be in the business of forecasting probabilities - they've already demonstrated that they can't, and they're not investors.  They should just be always staring into the abyss.

Of course the Federal Reserve doesn't read this blog, so far as I know.  But it's the sort of thing that doesn't require a majority vote for individuals to use in their personal lives.

21